Corpus Linguistics as a Tool for Metapragmatics in Japan

Language change has always reflected transformations of socio-cultural realities. However, in modern Japan, change in ‘the Japanese language’ in its conception as a monolithic vehicle of Japaneseness has been frequently perceived as a deterioration of linguistic substance, and by extension, as an erosion of order and culture. In this paper, software-based corpus linguistics methodology is applied to a corpus of newspaper articles within the framework of discourse analysis, with the aim of describing discourse actors and extracting pragmatic idiosyncrasies of the newspaper-mediated public metalinguistic discourse centred on language decline. My findings suggest that several pragmemes can be correlated with one or more of the main groups of discourse actors. These include the use of symbolic language, implications, objectifying language, and the construction of change as something happening (only) in the present.


Introduction
When Japanese sociolinguist Yonekawa Akihiko wryly pointed out that "for a person of culture, it is practically common courtesy to complain about the decline of language, to bemoan its sorry state and to declare one's worries about it (or at least to pretend being worried)" (Yonekawa 2002, 70), he skil-fully encapsulated the essence of a broad discourse spanning more than one hundred years, most outlets of mass media and many areas of society in modern Japan. The decline of language is one of the central topoi of post-war Japanese language criticism. Kotoba no midare, the focal term in this discourse, has been in use for many decades. Being a catch-all term for whatever rubs someone's linguistic sensibilities the wrong way, its essence has remained unchanged: language that is considered good, aesthetically pleasing and correct is threatened by elements which are seen as being at odds with this "good language".
While some discourses of aesthetics and questions of good or bad remain the domain of intellectuals and have little bearing on day-today life, language criticism is different. Philologist Gerd Antos argues that people are mainly interested in areas of language and communication that they feel are either fun or problematic (1996,3). Since every native speaker, having been born and raised in an environment saturated in their mother tongue(s) can be considered to be an expert in their native language(s), it seems natural to assume that, notwithstanding their level of education, they can and indeed often will have an opinion on what makes good or bad language. Considering that the permanent use of a common language can be seen as a condition for the permanent constitution of a shared social reality (Keller 2013, 26), it is not difficult to see why anything that seems to threaten this permanence will quite inevitably be met with criticism. This paper will focus on the "problematic" part of Antos' assertion, which can often be found to be related to manifestations of language change. Using corpus linguistics on a purpose-built corpus of newspaper articles, I will try to shed light on the question of what is seen as problematic about language (change), and how metalinguistic commentary is constructed.

2
Discourse (Analysis) and Metapragmatics processes of social construction, circulation and mediation of interpretative schemes, patterns of legitimisation and modes of action" (2013,46). I will follow Gardt's and Keller's understanding of discourse in this paper. True to the etymology of (meta)pragmatics, language criticism is concerned with actual language practice in a sociocultural context (cf. Silverstein 1993, 34). Analyzing social discourse can be seen as having an inherently metapragmatic side; discourse is shaped by language acts with specific goals, which themselves may and indeed often do become the subject of contemplation. In a similar way as "discourse", the term "metapragmatics" invites a number of more or less divergent readings (cf. Silverstein 1993;Mey 2001;Bublitz, Hübler 2007, 1). In this paper, I will use the term in the sense of pragmatics of metalanguage, along the lines of John A. Lucy's reflexive language. Lucy maintains that "metalinguistic activity is [...] fundamentally metapragmatic" (1993,17), as reflective activity will typically deal with the appropriate use of language in specific contexts of communication. Haugh (2018) demonstrates the usefulness of corpus-based metapragmatics within the wider scope of conversation analysis, but I would like to go a step further and connect corpus linguistics with metapragmatics and (non-linguistic) discourse analysis. This is broadly similar to Wright and Brookes' (2019) corpus-assisted study of right-wing British media and Saft and Ohara's (2006) study of Japanese newspaper editorials, but differs in terms of methodology and aim. Saft and Ohara use CDA as their practical framework, although the actual makeup of the corpus and the methodology used to analyze the corpus remain largely nebulous. Wright and Brookes explain their approach in detail, but use only basic corpus linguistics methodology since they are more interested in topoi than pragmemes, which I will focus on in this paper.
Pragmemes, to extrapolate from Capone (2005Capone ( , 1357, are speech acts with the goal to shape social reality within the restraints of sociocultural context, essentially signifying a structure that is "intrinsically made up of language, culture, and society" (Capone 2016, xvi-ii; see Mey 2016a for an in-depth discussion of the term). Pragmatics approaches language use from a holistic viewpoint, encompassing language embedded in its social and cultural surroundings in a way that a purely linguistic approach cannot cover (Senft 2014, 3-4). I believe this is essentially where discourse analysis, with its focus on untangling the knot of involved sociocultural norms and ideas, motivations and actions, intersects with (meta)pragmatics.
I have found the structured and accessible approach of Siegfried Jäger's critical discourse analysis (Jäger 2015) to lend itself very well to a basic structure for a corpus-assisted discourse study, and the outcome may go some way towards what Jacob L. Mey calls the "re-sponsibility of pragmatics seen through a societal prism", consisting in "'unveiling' cases of linguistic manipulation." (Mey 2016b, 128).

3
The Japanese Discourse on Language Decline Kotoba no midare can be loosely translated as "language [falling into] disarray", but for all intents and purposes, it denotes a decline of language. 1 While there is a case to be made for a pre-discourse within the broader debate on what the new, standardized language should be like, and therefore also what constitutes linguistic behaviour that goes against the grain of these newly-found ideals (i.e. debate about dialects as an impediment to national unity [hōgen ronsō; see i.a. Heinrich 2012, Seidl 2010, or the moral panic surrounding young women's deviant speech in the 1930s [cf. Inoue 2006, 37-41]), it was not before the 1950s that a public, broad metalinguistic discourse facilitated by newspapers and magazines emerged. While during the first two decades this discourse was to a large extent concerned with orthography reforms and other regulations, the 1950s also saw the spread of (kotoba/nihongo no) midare as a central term of language criticism in professional journals, and with the protocol of the 37th session of the Kokugo shingikai 2 (KSK) 1958, also a first mention in a document related to language policy. The term and its variations are increasingly visible in the KSK protocols from the 1970s onwards, and in the protocols of the 13th KSK (108th session, 1978) it is debated whether the current midare is to be understood as a call for action. This means that by the end of the 1970s, language decline had apparently become an issue of political consideration.

Building the Corpus
The discourse on the decline of languages can be observed to have become steadily more distinct during the 1970s. A thorough scouring of the Asahi and Yomiuri shinbun (the two largest Japanese newspapers) data archives returned 230 articles that refer to midare, its synonyms, or phenomena discussed as manifestations of midare in either headlines or the database-intern tags between 1945 and 1983. The corpus used for research was built iteratively by adding keywords gathered from the results of queries for headlines containing kokugo and nihongo in the Asahi Shinbun, which also served as a base to assess the proportion of articles dealing with language criticism within the larger frame of metalinguistic discourse. Additionally, specific terms that came up while analyzing the corpus later were also used for database searches. Two corpora were built; one consisting of pre-digital age articles (1945 to ca. 1983) in the form of scanned pages, the other consisting of digital text (1983-2019). 3 In this paper, I will focus on the part of data that was available as digital text. Since full text searches could be executed for these articles, additional query strings with Boolean operators were used.
The final corpus consists of 1,121 articles (623,320 tokens; Asahi and Yomiuri combined). The full text of these articles was copied and pasted in chronological order into a plain-text UTF-8 encoded file. Unnecessary metadata (header items: category, issue type etc.) was removed using regular expressions in a text editor. Other irrelevant text data was also deleted to forestall statistical distortion, i.e. irrelevant parts of multi-feature articles whose other parts were not connected to the research subject. The combined corpus 1945-2019 can be seen in graph 1, with both newspapers contributing roughly the same amount of articles: The steep increase from 1985 to around 2005 was found to mirror a general increase of nihongo or kokugo (the Japanese language) in the headlines, with nihongo taking the far larger share. Several factors come to mind: Apart from the introduction of CTS in the mid-1980s potentially leading to more articles in newspapers (Shiratori 2005, 510), a large part of nihongo-related articles cover Japanese as a foreign language (funding, promotion abroad, speech contests etc.), and the popularity of the Japanese language abroad. Additionally, an increasing number of articles deal with topics related to the spoken language (nihongo) from the 1990s onwards. This includes a large number of articles dealing with kotoba no midare. The decline of articles in the last decade can be regarded as a cooling of the inflated discourse of the 1990s, which was arguably fuelled by several factors, including language policy, business interests, and a phase of return to traditional values in the face of economic decline and the looming turn of the millennium.

Workflow Wrap-up
Corpus linguistics, i.e. analyzing large amounts of data with a focus on language patterns, can be a valuable tool for sociolinguistics and (meta-)pragmatics. Starting from the observation that "social acts lead to statistically salient language patterns" (Bubendorfer 2009, 2), I tried to identify such patterns through statistical analysis in order to uncover assumptions regarding language practice, languagemediated symbols and pragmemes used in conjunction with topics, timeframes and discourse voices. For analysis in this research, the program KhCoder was used, which was developed and is actively maintained by sociologist Higuchi Kōichi at Ritsumeikan University. 4 KhCoder was developed to analyze large corpora of Japanese or English 5 text, using morphological sequencer engines Chasen and Mecab, and the free R library of statistical computing as a base for many of its tools. The workflow described here is essentially inductive; using various tools, conspicuous word clusters and correlations are inspected, and working assumptions are built. These are then formalized as codes and tested on the corpus.
Before proceeding with the actual analysis, statistically salient ngrams were identified via the Word Clusters module and added as user-defined lemmata. The corpus was then tagged, i.e. through the use of xml tags, units of text were circumscribed for ease of analysis, specifically as decades, years, and single articles. In this paper, articles as units of analysis will be marked by adding [a] to the description or result of the operation; [p] for paragraphs, [s] for sentences.
The first step of analysis was to generate a list containing all tokens arranged by POS-type in order of term frequency. Words with an unexpectedly high frequency and words that seemed out of place in the corpus were noted for subsequent analysis. For a first overview of conspicuous word clusters, a co-occurrence network (CON) was generated based on the Jaccard coefficient. 6 While Jaccard is commonly used in corpus linguistics, it has the inherent disadvantage of lacking a clear-cut way to determine significance thresholds (Biruhs 2010, 31), which have to be determined heuristically (Moore 2004, 333). Jaccard coefficients in this paper are therefore not illustrative as, for instance, percentages or chi-squared values, and are supported by explanation and interpretation of context.
KhCoder allows for the use of codes, i.e. containers for arbitrary collections and combinations of tokens that can be combined using a syntax of Boolean and other logical operators. Analysis via coding allows for rather abstract conditions to be formulated and was employed frequently, though some caution is required. For example, if a correlation between two codes is determined, this only signifies that two codes are present within the same corpus segment. They would not necessarily have to be connected. To alleviate this problem, care was taken to train all codes on a manually selected set, adjusting the codes as needed for better efficiency and eliminating false positives as far as possible. Furthermore, methodological triangulation in Kh-Coder (predominantly KWIC searches, hierarchical cluster analysis, word association analysis [WAA] and code correspondence analysis) was used to fine-tune codes and confirm findings.

6
Pinpointing Pragmemes Media texts are a treasure trove of pragmemes and indicators of metapragmatic awareness; a qualitative, critical approach will readily reveal devices like rhetorical questions, categorical assertions, hedging etc. along which implications, propositions or presuppositions (Hardman 2010, 263;Overstreet 2010, 267;cf. Wright and Brookes 2019). While I decided to focus on statistically salient pragmemes, this does not mean I completely dispensed with (con)textual qualitative examination of texts. Rather, focused cases served as illustrative examples, and, more importantly, helped formulate working assumptions. There were found to be three distinct actor groups involved in shaping the discourse. These are editors, readers, and experts. While it is true that ultimately all text in the corpus is likely to have passed through the hands of newspaper editors, the code *Editors (5.6% of all articles) encompasses only articles that are designated as editorials or opinion pieces, which appear as columns with distinct names and are typically situated in a prominent spot among the first pages of the paper. They are thought to represent the official stance of the newspaper and can be expected to be consciously penned with the goal of influencing public opinion (Saft, Ohara 2006, 85). *Readers (30.7%) contains all articles that are either clearly designated as letters to the editor or that were determined to consist largely of reader/non-editorial input, e.g. small-scale, largely uncommented manon-the-street interviews conducted by the paper. The third category *Experts (21.4%) describes articles containing references to "experts" on linguistic matters, i.e. linguists, playwrights, lyricists, stage actors etc. Articles in *Experts were found to often take the form of an interview, a guest commentary, or to be part of a multi-feature piece written by one or several experts. Articles that could not be clearly connected to any of these three types were grouped as *Misc.

Living Language
Likening language to an organism is not new symbolism; it can, for example, be found in 19 th century German critical literature (Heringer, Wimmer 2015, 63). This "organic view" of language can be easily expressed in a code, which was found in 5% of all articles and appears in all decades. While most instances use the direct metaphor of a living creature (ikimono), a smaller number encapsulated the same image in a clause (e.g. "since language is alive, …").
There is a strong correlation between *Organism and words and phrases that are used to indicate change or transformation (henka, kawaru, utsurikawaru etc.). *Change occurs in 33% of all sentences which contain *Organism. This can be read as acknowledgement of the processual nature of language in a society: "Living things must always evolve. Therefore, language must change, too." (YS 2003-06-05).
However, this naturalistic view of language change is not necessarily always used in a straightforward, accepting way. About a third of all instances of the code are followed by a concessive clause used to put this view into perspective: even if change is not to be avoided, it may not be wished for, especially if personal linguistic sensibilities are offended: • "Language is a living organism and it may well change over time, but simply exchanging reru for rareru would be very problematic" (AS 1992-11-02);

Bernhard Seidl Corpus Linguistics as a Tool for Metapragmatics in Japan
Ca' Foscari Japanese Studies 13 | 1 149 European Approaches to Japanese Language and Linguistics, 141-164 • "Personally, I believe that language is alive and ever in transition. But even I cannot help but be concerned about how zenzen is used in zenzen ii" (AS 1991-08-29); • "Language is a living thing. It changes over time. Nonetheless, we must not lose sight of the [connection between language and] our history and customs" (AS 1986-03-08).
If language is a "living thing", it would seem natural to find instances of comparing kotoba no midare to death or dying, but this extension of the organic metaphor is almost non-existent in the corpus. Likewise, there are only a handful of instances that liken the state of language (use) to disease. These typically appear only once or twice and it remains contextually unclear whether this disease is of the language or of the speaker; e.g. tachiageru-byō (a "disease" characterized by the use of the verb tachiageru in various contexts thought to be inappropriate) or riron fumei shōkōgun ("illogicality syndrome", the use of syntactically incoherent sentences).

Foreign Floods
The most conspicuous metaphor found in this corpus is the image of a torrent or flood (hanran; occasionally also zōsui or kansui) in the sense of an inundation of unwelcome linguistic elements. In a similar way as with the organic metaphor, this plays into Link's notion of a collective symbol; indeed, flood (Flut) is an example Link gives for a symbol with an inherently negative connotation (Jäger 2015, 62). The commonly understood implications of such symbols are what makes them useful; for example, a flood will damage carefully cultivated fields and human infrastructure and, being a natural disaster, there is precious little that people can do about it. To put it slightly more dramatically, it carries with it the image of being at the mercy of an existential threat: "Does this torrent of katakana-words not mean chaos and calamity for the Japanese people and their culture?" (AS 2000-03-07). *Flood is found in 9.5% of all articles and is almost exclusively associated with *Foreign_words (J[p]=0.14; cf. the next best match *Officialese 7 J[p]=0.05). A WAA limited to nouns confirms this and shows that the most statistically salient collocators for *Flood are katakanago, gairaigo, gaikokugo (predominantly used to denote foreign words not written in katakana), rōmaji and yokomoji, all of which were found to be used largely synonymously in the corpus and could therefore be collected in a code. The frequency of this code can be seen in graph 2.
7 Language used in public administration; described as convoluted and inundated with loanwords. Testing via code correlation confirms that approximately two thirds of all instances of *Flood are concerned with these topics. The rest is rather mixed and includes a variety of topics (buzzwords, incorrect keigo, exclamation markers in Japanese text, or youth language in general).
Hanran, as an image for the threat of foreign words, seems to have emerged in the newspapers in the mid-1950s, and by the end of the 1980s gairaigo no hanran had been firmly entrenched as a stock phrase, to the point where it made its way into the 5 th session of KSK 17 (1988), where we can find a section regarding the "problem of what is referred to as flood [iwayuru hanran] of foreign words 8 and loanwords".
A closer look reveals the "foreign influence" on the Japanese language to be anglicisms; a WAA [p] shows eigo (English) as one of the top three collocations for *Foreign_Words (J=0.09), the other two being nihongo (J=0.11) and iikae ("rephrasing"; J =0.1).
Iikae bears closer examination: it is a keyword in the sub-discourse on foreign language/words in Japanese in the 2000s, and may be seen as an effect of the flood pragmeme. It is hardly surprising that a loud enough warning call against a flood was answered with efforts 8 Gaikokugo, which is semantically ambiguous, since the suffix -go can denote a language or a word.

Bernhard Seidl
Corpus Linguistics as a Tool for Metapragmatics in Japan to build dams, to stay in the realm of metaphors. Indeed, when Culture Secretary Tōyama Atsuko commissioned the National Institute for Japanese Language and Linguistics (NINJAL) in 2002 to form a committee on the question of foreign words/loanwords in Japanese, with an expressed goal of finding Japanese 9 alternatives for katakana-words that were deemed difficult to understand, this can be seen as an effort to build a dam for protection against the flood. This was certainly fueled by the severe lashing the newspapers had been giving the katakana-inundated "officialese" since the late 1990s.
The project with the bulky title "Proposal for exchanging foreign words in favour of beautiful Japanese [words]" (Gairaigo ni kawaru utsukushii nihongo no teian) was launched with the immediate goal of dealing with this "officialese" (YS 2002-06-26), and both Yomiuri and Asahi regularly covered the committee's actions and word lists. In a 2004 survey conducted by the NINJAL, 61.3% of the surveyed people were found to be supportive of initiatives like the Gairaigo iikae teian, but at the same time, the majority of the respondents also said they were in favour of katakana words rather than Japanese words in the cases of unwieldy words (e.g. kōkyōshokugyōanteijo, employment office → harō wāku ["Hello work"]) or expressions with negative connotations (e.g. rōjin, elderly [person] → shirubā ["silver"]) (NINJAL 2004).
This hints at a certain ambivalence towards katakana-words, which is also visible in the large number of relevant dictionaries and books published in the 1990s and 2000s. On one hand, this confirms the view of katakana as something of an obstacle to communication, but, on the other hand, many of these books have a distinctively playful makeup and their titles and blurbs often imply the prospect of an increase in social prestige for being "in the know".
Notwithstanding this ambivalence, the discourse is clearly dominated by the critical voices. The following examples are quite representative of the sentiments that are particularly pronounced in *Readers: • "Especially problematic is the excessive use of katakana-words.
Everything is katakana-English nowadays: song titles, product names, even the names of government initiatives" (YS 1996-05-08); • "Strange pseudo-English everywhere, incomprehensible Latin letters and katakana wherever you look: Many people can't even speak their own mother tongue properly any more.
[…] Am I really in my home country, is this really still Japan?" (AS 1999-11-14); • "There is a flood of katakana words and abbreviations. It makes me sad to witness beautiful Japanese words disappear. I know that language is always fluctuating, but that should be even more reason to preserve our beautiful Japanese words." (YS 2013-04-23).
In the early 2000s, the government's push for the introduction of English into primary school curricula re-ignited the debate. Even radical right-wing politician and erstwhile writer Ishihara Shintarō, then Governor of Tōkyō, regaled the public with his opinion on the matter: "They haven't even mastered Japanese yet, and still they are supposed to learn a foreign language via Japanese? What's the point? You just need to take a quick look at the gibberish in the media to see that Japanese is going downhill fast" (AS 2006-04-08). While some sub-discourses and topics have proved to be rather ephemeral, the debate about foreign elements in the Japanese language can be observed to be very persistent. After many decades, as Stanlaw (2004, 1) succinctly puts it, it remains ultimately unclear to the observer how society actually does regard foreign elements in the Japanese language: as "[a] problem (Yokoi, 1973), an obstacle to communication (Hirai 1978), something 'stylish' (Kawasaki, 1981), or some kind of pollution (Kirkup, 1971;Morris, 1970)". Within the discourse on language deterioration, at least, "foreign words" seem to be one of the most popular scapegoats.

Objectivity of Statements
Another salient pragmeme is the use of passive verbs to imply objectivity or a general validity to the statement. Passive clauses in principle do not need a subject/agens, so it stands to reason that using the passive voice is used to convey a certain detachment of the speaker from the message; it can be understood as an expression carrying "[t]he nuance that this is some kind of expert or public opinion rather than the speaker's own." (Oyamada et al. 2012, 62). While this structure can be found in all types of articles in the corpus, statistically, the likelihood of article types containing it in the context of language criticism was highest for *Readers (J[a]=0.11) and *Experts (J[a]=0.1). *Misc were found to be less likely (J[a]=0.08) and *Editors rather unlikely (J[a]=0.02) to contain this kind of passive clause.
In *Readers, passive clauses are often used as an opener, as in the following examples (translated as passive for illustrative purposes and underlined): • "It has been said that the deterioration of our language has become intolerable. There are many things going on in our language lately that I find quite unacceptable" (AS 2005-09-10);

Bernhard Seidl Corpus Linguistics as a Tool for Metapragmatics in Japan
Ca' Foscari Japanese Studies 13 | 1 153 European Approaches to Japanese Language and Linguistics, 141-164 • "The decline of our language has been pointed out for some time now. Lately, however, not only the language young people use seems strange to me, but also that used on TV" (YS 2004-12-21); • "The deterioration of language among young people is talked about, but I would like to propose adding the rudeness of 'old people's language' to this" (AS 2018-06-17).
As can be seen from these instances, the use of passive is actually frequently followed by an explicitly subjective statement. It seems to me that this kind of opening statement helps facilitate the subjective statement following it. After all, it is not "I" who is complaining; rather the writer merely subscribes to a point of view that is implied to be general knowledge or common sense, and hence places the personal misgivings voiced in the following statement on solid ground.

Temporal Proximity
*Readers were found to have another feature much less present in other article types. A WAA [a] based on *Readers yields tadashii (correct), utsukushii (beautiful), saikin (lately), omou (to think, be of the opinion that), nihongo (Japanese), midare/midareru and terebi (TV) as 8 of the top 10 collocations (the remaining two are koe and kiryū, names of columns for letters to the editor). None of these words are a statistically salient collocation for other types of articles in the corpus.
Visualizing the results as a CON confirms that a direct reference to midare framed as something contemporary and put into contrast with how Japanese is supposed to be (correct, beautiful) can indeed be seen as a salient pragmeme of *Readers; even without allowing for synonyms, 45% of all instances where midare and saikin co-occur within one sentence are found in *Readers. This can be interpreted as an implicit form of argumentum ad antiquitatem: something is now different from how it used to be. If it has always been that way, that way must have been good. Changing things therefore must be bad.
Ultimately, this kind of linguistic conservatism is what drives the whole discourse by creating a sense of loss and urgency. Finding this pattern in the corpus also confirms Keller's observation that "[i]t is only ever the present state of the language that is threatened by decline and decay; there has never been a discourse on how 'deteriorated' or bad some historical language or variety was" (Keller 2004, 4). Language decline happens in the world that the critic lives in, and is asserted by contrasting the present state of the language with how it used to be (i.e. good, beautiful etc.), with no regard, it might be added, to actual history: current critics rarely reflect upon the fact that the same decline of language was likely lamented regarding their own generation and its language use, and that linguistic norms and rules were at no point in time so clear-cut as to allow for a simple verdict about right or wrong to be made in most cases. To paraphrase Heringer and Wimmer: many critics yearn for their language to revert to the state it was never in to begin with (2015, 15).

Unity of Language, Culture (and Nation?)
Upon observing a high frequency of words like culture and tradition via POS frequency lists, synonyms were gathered and factored into *Culture_Tradition, which was found to be present in 19.5% of all articles. *Change, *Language and *Ideal_ Japanese were found to be the strongest collocator codes [p]. This was taken to be indicative of an association between the concepts of culture and language, and using both codes on sentences yielded numerous examples to corroborate this assumption: • "Language is synonymous with the culture of a nation" (YS 2000-09-12); • "It is because you are using this kind of language that Japanese traditions are falling apart" (AS 2000-01-05); • "For me, patriotism means nurturing the culture of our beautiful Japanese language" (AS 2001-06-23); • "In order to preserve our culture, we must also preserve our language and see that it doesn't fall into decline" (YS 2016-03-29).
Further investigation yields interesting results indicating notions of community and continuity. The top collocators for *Culture_Tradition [s] are nihon (Japan), mamoru (preserve, save), gengo (language), nihonjin (Japanese [people]), kuni (nation), rekishi (history), shakai (society), taisetsu (important), utsukushii (beautiful), -iku (auxiliary verb). Taisetsu and -iku bear closer examination: the first is mostly used as an adverbial phrase taisetsu ni suru (to take care of, to cherish), the latter is used to express a sort of projected continuity, indicating actions that start in the present and will be or should be continued in the future. Together with mamoru and rekishi, -iku is used to construct an imagery of continuity. Constraining the results to adjectives, the results can be grouped into words describing necessity and/or care[fulness] (taisetsu, jūyō, daiji, hitsuyō), language ideals (tadashii, utsukushii, yutaka), youth and novelty/freshness (wakai, atarashii).
In summary, the concepts of language and culture often seem to be linked, which potentially elevates the discourse on language decline by some order of magnitude: if language and culture are intertwined, the deterioration of one must have an ill effect on the other, which, of course, has been a major argument of language critics everywhere. "Damaging" the language can and must be seen as a severe problem if it is believed to lead to a damage of cultural substance. In this way, an implication is made to the discourse participants that nurturing "proper" language is an ethical duty to one's nation or community. I will examine this pragmeme of "demand for action" separately.
This kind of reasoning is not without its problems in the light of the nationalist appropriation of kokugo (the national language) from the Meiji period to the end of WW2. For example, in Kokugo to wa nanzo ya (1942), Yamada Yoshio stresses the inseparable ties between nation, national spirit and national language, and argues that treating the national language irreverently is tantamount to damaging the national spirit and, therefore, ultimately the nation; further, the key work Kokugo no tame (1897) by Ueda Kazutoshi bears the motto "Kokugo is the bulwark of our imperial family / Kokugo is the loving mother of our people" on its first page.
While less direct, examples of this kind of identification of language with culture and nation can also be found in the corpus, as in this interview with writer Moriuchi Toshio in the Yomiuri: "We think about things and concepts using language. A deterioration of language therefore means a decline of culture, and ultimately a decline of our nation. This is a most serious issue" (2002-06-16). This is a slippery slope fallacy, further examples of which can be found scattered all over the corpus. Take, for instance, this Asahi editorial, which incidentally contains most of the pragmemes mentioned so far: Language is a living organism; it changes over time. Even so, we must cherish the connection between language, history and tradition. Take, for example, the flood of foreign words or the confusion about how they are written. Some people say this contributes to variety in the language. However, a laissez-faire stance will undoubtedly end up endangering communication, widening the gap between generations, and may ultimately even lead to the demise of the Japanese language. (AS 1986-03-08) Tessa Carrol argues that the often observed link between nation, order and language in many cases points to a certain affinity towards nihonjinron thought (2001,77), and Komori (2003) sees a clear link between the Japanese language boom of the 2000s and the rise of an inward-looking nationalism (also cf. Kayama Rika's concept of a "mini nationalism syndrome"). It has been argued that, in general, "language is a prime determinant of nationalist identity" (Billig 1995, 29 in Wright, Brookes 2019, and I think that this is quite visible in the discourse on language decline: while upholding the unspoken premise that there is only one Japanese language rather than the multitude of varieties and indeed languages (cf. the Ryūkyūan languages) that together make up a rich and varied garden, Japanese is used as an implicit symbol for notions of national unity and uniformity. This is by no means limited to newspapers; on the contrary, it is my impression that this kind of thinking is much more obvious and common in the language-critical popular literature, where a yearning for the beautiful and now lost "language of yore" can be read as a bemoaning of a better Japan long lost, or as a fear of losing cultural identity: haplessly swept away in a torrent of foreign words or youth language, as it were.

Demand for Action
There are numerous ways in which a demand for action can be expressed in Japanese, but for this corpus, the following patterns could be determined: Tai-form (volitional) to express a desire for action to be taken, hitsuyō da/ga aru to denote the necessity of action, and finally nakereba/nakute wa ikenai/naranai and the suffix beshi to indicate that something ought to be done. These can be understood to differ not in essence, but in intensity, conveying varying levels of urgency (Saft, Ohara 2006, 96-7). While the distinction between more or less urgency is surely a worthwhile one to make in Saft and Ohara's research on the pursuit of militarism in Japan as reflected in the media, I decided this to be of less importance for the language-critical discourse. Therefore, all these expressions were combined as (demand for) *Action.
This code was found in 57% of all articles in the corpus, but could not be clearly connected to any of the actor groups. A demand for action can therefore, rather unsurprisingly, be seen as a general characteristic of the language-critical discourse. The object or audience of demand can again be gathered from a combination of WAA and CO of codes [a,p]. *Youth and *School (both J[p]=0.08) were found to be the concepts most distinctly associated with *Action; *Polite_language and *Foreign_words (both J[p]=0.07) were the most related topics. Simply put, demands here can be expected to concern the language use of young people and/or the significance and shortcomings of kokugo (the school subject Japanese) on one hand, and the need to address the problem of foreign words, buzzwords and neologisms on the other. This confirms the centrality of young people as discourse actors without a voice, and of topics related to semantic change as the most visible manifestation of language change. Because of this visibility and the immediacy felt when confronted with (new) words one does not understand, semantic change is therefore also the most likely target of language criticism, and subsequently a demand for action to be taken to hem in the detrimental effect on the Japanese language is to be expected.
Interestingly, *Action was not found to necessarily correlate with the visibility of a topos in the discourse; for example, the question of ra-nuki kotoba (the practice of dropping /ra/ in the potential forms of certain verbs) is one of the most long-lived and pervasive topics in the corpus, with 21% of all articles containing some reference to it, but it could not be found to correlate significantly with *Action (J[p]=0.03).

Ideal Language
Even in language communities with a standard language, there is no such thing as universally accepted, objective criteria of "good" or aesthetically pleasing language. Still, it is precisely the notion of "bad language (practice)" that is perceived as a threat to the integrity of the language (community). It is therefore interesting to examine how good and bad language is constructed in the discourse. Two general practices can be discerned that I will investigate further in the next two sections: circumscribing good language indirectly by describing bad language, and defining desirable language directly through its attributes. Both can be approached through WAA aided by KWIC analysis for clarifying context when necessary, and subsequent grouping.
All of these categories consist of a variety of topoi, and mapping the most prevalent ones on a timeline shows the short lifespan of some and the perseverance of others. The rise and fall of some of them can be seen to reflect changes in society: the sub-discourse on women's language, for example, was a central issue in the late sixties and seventies, with its own topics and pragmemes, but the focus slowly changed from the decline of women's language (seventies to early nineties) to a measure of grudging acceptance of the changing reality of gender images (mid-nineties to mid-2000s), and has all but disappeared by 2019. While I have already touched upon the topic of foreign words, I would like to examine the category of semantics closer, as lexical changes are typically the most easily noticed and therefore most readily used to symbolize bad language practice.
Semantic change was found to be mostly associated with youth (language), and focused on neologisms (shingo), buzzwords (ryūkōgo) and abbreviated words (ryakugo), with the combined code *New_ Words having the highest frequency of all codes describing a thematic category (43.4% of all articles). In essence, the sub-discourse staked out by these terms revolves around mostly short-lived words that are claimed to be incomprehensible and/or unacceptable. The often equally short-lived, but sometimes intense attention some of these words receive is reminiscent of a (small-scale) moral panic, with a sudden explosion of comments that are quick to extrapolate from the words to their users, and from linguistic deviance to moral decline, before ceasing as abruptly as the use of the offending word(s) themselves. Incidentally, this is not limited to topics linked to lexicality and can be observed with a variety of topics in the discourse on kotoba no midare.
It is noticeable that *Experts tend to explain youth language in a more positive light. This observation can be somewhat substantiated by combining *Experts with *Youth_language / *New_words into one code and examining its co-occurrence with *Negative_assessment, which tries to circumscribe the various forms of negative, disparaging or concerned commentary found in the corpus. While the correlation of the combined code with *Negative_Assessment is highest [a] with *Editorials (J=0.16), and lower for *Readers (J=0.12) and *Others (J=0.09), the correlation for *Experts is lowest of all (J=0.06). The following statement by philologist Koyano Tetsu in the Asahi Shinbun is an explicit example for this tendency: Young people always keep making up new words. They possess the freedom needed for a playful use of language. This kind of language is typically used only when they are with their friends, so I don't see why people make such a fuss about it. In fact, I rather think this only leads to young people losing their power to break rigid patterns, and saps the vitality of language. (AS 1999-03-11b)

Direct Definition
The shape of ideal Japanese as an antithesis to midare can be gleaned from comparing the results of a WAA of *Japanese and *Language_ Decline [s] and limiting the results to adjectives, which gives beautiful (utsukushii, kirei), correct (tadashii), and rich/colourful (yutaka) as the top collocates used to positively describe language. In 35% of the corpus, at least one of these is used to describe Japanese, with 8.7% of all articles containing a combination of two or all three, and "beautiful" the most frequent. Expressing these qualities as codes and comparing their co-occurrence [a] with that of other codes which were compiled in the course of this project hints at who these qualities are mostly demanded from: *Youth (J=0.3), *Mass_ media (J=0.29), *Women (J=0.23) and *Education (J=0.2).
While correctness may in theory be argued by appealing to a higher authority (norms established through language policies, exegetical literature 10 and experts), "beautiful" and "rich" are every bit as problematic as "deterioration" due to the inherently subjective nature of these terms: pointing to language norms is of little use here, and so the question arises as to what exactly language critics mean when they use them. There are only a few articles in which examples are given, and these are far removed from day-to-day communication (traditional poems, storytelling arts, Japanese classics), which is typically the bone of contention.
A WA/KWIC analysis of *Beautiful_language [p/s], limited to *Readers (found to have the highest prevalence of sentences describing desired language), reveals which specific qualities are presumably meant: A fluent/eloquent (sawayaka), gentle/kind (yasashii; egao), rich/colorful (yutaka), and/or correct (tadashii) language having a pleasant (kokochiyoi; utsukushii) sound (hibiki), which has to be strived for (kokorogakeru), or which needs to be preserved (mamoru; taisetsu ni suru) because of its close association with culture (bunka).
Thus, inherently subjective attributes are credited with a high argumentative value, typically without bothering to explain the rationale for this judgment: language is beautiful or ugly, correct or incorrect because the language critics say it is. This kind of ambiguity can also be found in official documents describing language in the context of language preservation policies. The KSK, for instance, described an ideal Japanese language in a working paper (Kokugo kyōiku no shinkō ni tsuite, 1972) as heimei (clear, unambiguous), tekikaku (precise), utsukushii, and yutaka. Later institutions of language policy used an almost identical array of qualities to describe language considered beneficial, also without clarification.

7
Concluding Remarks Implicitness is sometimes seen as a characteristic feature of pragmatic studies (Jacobs 1999, 5) and typical of media texts (Hardman 2010, 264), and it was indeed observed to be central to most pragmemes in the corpus: the meaning of the structures I described does not end with 10 Selling "correct Japanese" through books, language certificates for native speakers, TV shows and video games has been a big market segment since the late 1990s.
conveying certain information. Rather, they carry with them a wealth of implications. In a similar way, metaphors have been described as a "paradigmatically pragmatic phenomenon" (Camp 2010, 264), and were also found to be distinctly visible in the corpus. Drawing on Jürgen Link's concept of collective symbols (Link 2012, 135-7;Jäger 2015, 60-1), I regard metaphors in the context of social discourse as more than simply a stylistic choice. Imagery such as a 'flood' of foreign words represent a slice of "the entirety of what may be called the 'imagery' of a culture; the entirety of a culture's most widely understood allegories, emblems, metaphors, exemplary cases, and illustrative patterns" (Link 1997, 25). These symbolic cues for emotions can be expected to be subconsciously picked up by the intended audience, and therefore also to have a distinct influence on the language critical discourse. On a more comprehensive level, it is neither its existence nor its core tenets that sets the Japanese discourse apart from that of other nations. The question of language decline is not a unique characteristic of Japanese metalinguistic discourse; it can be expected to be found in some form in most nations with a normative standard variety. What makes the Japanese discourse somewhat of a special case, then, is its size and extent. While I have sketched some of it briefly, it goes well beyond the scope of this paper to describe the foundations of this Japanese fascination with the Japanese language, but, as early as 1942, linguist Ishiguro Yoshimi noted that "[the] last years brought a large number of articles in newspapers and magazines concerning the problems of our language, and there are constantly new books being published. This shows the great interest of the people in matters of the language" (Ishiguro 1942, 53).
More recently, million-sellers like Nihongo renshūchō (1999), Koe ni dashite yomitai nihongo (2001) or Jōshiki toshite shitte okitai nihongo (2002) are only the professional tip of an iceberg of "amateur linguistics": books, magazines and other media explaining the peculiarities, weal and woe of their mother tongue to the Japanese. To an extent this can be seen as a manifestation of the language-critical discourse in a society which has been perpetuating the narrative of a widespread decline of the language, but the same discourse also ties in a broad multi-media narrative of Japanese as a symbol and vehicle of culture, or "Japaneseness". In fact, I would argue that the self-exoticizing and sometimes outright nationalistic nihonbunkaronliterature intersects considerably with the broader Japanese metalinguistic discourse.
From a more sociolinguistic point of view, language criticism in the sense of a negative outlook on language change can be regarded as a natural side effect of language standardization and the determination of linguistic norms. But standard language does not exist in a cultural or social vacuum, and will therefore be subject to societal stress caused by change at some point, invariably leading to